tipc: make subscriber server support net namespace

TIPC establishes one subscriber server which allows users to subscribe
their interesting name service status. After tipc supports namespace,
one dedicated tipc stack instance is created for each namespace, and
each instance can be deemed as one independent TIPC node. As a result,
subscriber server must be built for each namespace.

Signed-off-by: Ying Xue <[e-mail]>
Tested-by: Tero Aho <[e-mail]>
Reviewed-by: Jon Maloy <[e-mail]>
Signed-off-by: David S. Miller <[e-mail]>

net/tipc/core.c

···
 	err = tipc_nametbl_init(net);
 	if (err)
 		goto out_nametbl;
+
+	err = tipc_subscr_start(net);
+	if (err)
+		goto out_subscr;
 	return 0;
 
+out_subscr:
+	tipc_nametbl_stop(net);
 out_nametbl:
 	tipc_sk_rht_destroy(net);
 out_sk_rht:
···
 
 static void __net_exit tipc_exit_net(struct net *net)
 {
+	tipc_subscr_stop(net);
 	tipc_net_stop(net);
 	tipc_nametbl_stop(net);
 	tipc_sk_rht_destroy(net);
···
 
 	get_random_bytes(&tipc_random, sizeof(tipc_random));
 
-	err = register_pernet_subsys(&tipc_net_ops);
-	if (err)
-		goto out_pernet;
-
 	err = tipc_netlink_start();
 	if (err)
 		goto out_netlink;
···
 	if (err)
 		goto out_sysctl;
 
-	err = tipc_subscr_start();
+	err = register_pernet_subsys(&tipc_net_ops);
 	if (err)
-		goto out_subscr;
+		goto out_pernet;
 
 	err = tipc_bearer_setup();
 	if (err)
···
 	pr_info("Started in single node mode\n");
 	return 0;
 out_bearer:
-	tipc_subscr_stop();
-out_subscr:
+	unregister_pernet_subsys(&tipc_net_ops);
+out_pernet:
 	tipc_unregister_sysctl();
 out_sysctl:
 	tipc_socket_stop();
 out_socket:
 	tipc_netlink_stop();
 out_netlink:
-	unregister_pernet_subsys(&tipc_net_ops);
-out_pernet:
 	pr_err("Unable to start in single node mode\n");
 	return err;
 }
 
 static void __exit tipc_exit(void)
 {
-	unregister_pernet_subsys(&tipc_net_ops);
 	tipc_bearer_cleanup();
 	tipc_netlink_stop();
-	tipc_subscr_stop();
 	tipc_socket_stop();
 	tipc_unregister_sysctl();
+	unregister_pernet_subsys(&tipc_net_ops);
 
 	pr_info("Deactivated\n");
 }

net/tipc/core.h

···
 	/* Name table */
 	spinlock_t nametbl_lock;
 	struct name_table *nametbl;
+
+	/* Topology subscription server */
+	struct tipc_server *topsrv;
+	atomic_t subscription_count;
 };
 
 #ifdef CONFIG_SYSCTL

net/tipc/server.c

···
 	struct socket *sock = NULL;
 	int ret;
 
-	ret = tipc_sock_create_local(s->type, &sock);
+	ret = tipc_sock_create_local(s->net, s->type, &sock);
 	if (ret < 0)
 		return NULL;
 	ret = kernel_setsockopt(sock, SOL_TIPC, TIPC_IMPORTANCE,

net/tipc/server.h

···
  * @conn_idr: identifier set of connection
  * @idr_lock: protect the connection identifier set
  * @idr_in_use: amount of allocated identifier entry
+ * @net: network namspace instance
  * @rcvbuf_cache: memory cache of server receive buffer
  * @rcv_wq: receive workqueue
  * @send_wq: send workqueue
···
 	struct idr conn_idr;
 	spinlock_t idr_lock;
 	int idr_in_use;
+	struct net *net;
 	struct kmem_cache *rcvbuf_cache;
 	struct workqueue_struct *rcv_wq;
 	struct workqueue_struct *send_wq;
···
 				  struct sockaddr_tipc *addr, void *usr_data,
 				  void *buf, size_t len);
 	struct sockaddr_tipc *saddr;
-	const char name[TIPC_SERVER_NAME_LEN];
+	char name[TIPC_SERVER_NAME_LEN];
 	int imp;
 	int type;
 };

net/tipc/socket.c

···
  *
  * Returns 0 on success, errno otherwise
  */
-int tipc_sock_create_local(int type, struct socket **res)
+int tipc_sock_create_local(struct net *net, int type, struct socket **res)
 {
 	int rc;
 
···
 		pr_err("Failed to create kernel socket\n");
 		return rc;
 	}
-	tipc_sk_create(&init_net, *res, 0, 1);
+	tipc_sk_create(net, *res, 0, 1);
 
 	return 0;
 }

net/tipc/socket.h

···
 
 int tipc_socket_init(void);
 void tipc_socket_stop(void);
-int tipc_sock_create_local(int type, struct socket **res);
+int tipc_sock_create_local(struct net *net, int type, struct socket **res);
 void tipc_sock_release_local(struct socket *sock);
 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
 			   int flags);

net/tipc/subscr.c

···
 	struct list_head subscription_list;
 };
 
-static void subscr_conn_msg_event(struct net *net, int conid,
-				  struct sockaddr_tipc *addr, void *usr_data,
-				  void *buf, size_t len);
-static void *subscr_named_msg_event(int conid);
-static void subscr_conn_shutdown_event(int conid, void *usr_data);
-
-static atomic_t subscription_count = ATOMIC_INIT(0);
-
-static struct sockaddr_tipc topsrv_addr __read_mostly = {
-	.family			= AF_TIPC,
-	.addrtype		= TIPC_ADDR_NAMESEQ,
-	.addr.nameseq.type	= TIPC_TOP_SRV,
-	.addr.nameseq.lower	= TIPC_TOP_SRV,
-	.addr.nameseq.upper	= TIPC_TOP_SRV,
-	.scope			= TIPC_NODE_SCOPE
-};
-
-static struct tipc_server topsrv __read_mostly = {
-	.saddr			= &topsrv_addr,
-	.imp			= TIPC_CRITICAL_IMPORTANCE,
-	.type			= SOCK_SEQPACKET,
-	.max_rcvbuf_size	= sizeof(struct tipc_subscr),
-	.name			= "topology_server",
-	.tipc_conn_recvmsg	= subscr_conn_msg_event,
-	.tipc_conn_new		= subscr_named_msg_event,
-	.tipc_conn_shutdown	= subscr_conn_shutdown_event,
-};
-
 /**
  * htohl - convert value to endianness used by destination
  * @in: value to convert
···
 			      u32 found_upper, u32 event, u32 port_ref,
 			      u32 node)
 {
+	struct tipc_net *tn = net_generic(sub->net, tipc_net_id);
 	struct tipc_subscriber *subscriber = sub->subscriber;
 	struct kvec msg_sect;
 
···
 	sub->evt.found_upper = htohl(found_upper, sub->swap);
 	sub->evt.port.ref = htohl(port_ref, sub->swap);
 	sub->evt.port.node = htohl(node, sub->swap);
-	tipc_conn_sendmsg(&topsrv, subscriber->conid, NULL, msg_sect.iov_base,
-			  msg_sect.iov_len);
+	tipc_conn_sendmsg(tn->topsrv, subscriber->conid, NULL,
+			  msg_sect.iov_base, msg_sect.iov_len);
 }
 
 /**
···
 {
 	struct tipc_subscription *sub = (struct tipc_subscription *)data;
 	struct tipc_subscriber *subscriber = sub->subscriber;
+	struct tipc_net *tn = net_generic(sub->net, tipc_net_id);
 
 	/* The spin lock per subscriber is used to protect its members */
 	spin_lock_bh(&subscriber->lock);
···
 
 	/* Now destroy subscription */
 	kfree(sub);
-	atomic_dec(&subscription_count);
+	atomic_dec(&tn->subscription_count);
 }
 
 /**
···
  */
 static void subscr_del(struct tipc_subscription *sub)
 {
+	struct tipc_net *tn = net_generic(sub->net, tipc_net_id);
+
 	tipc_nametbl_unsubscribe(sub);
 	list_del(&sub->subscription_list);
 	kfree(sub);
-	atomic_dec(&subscription_count);
+	atomic_dec(&tn->subscription_count);
 }
 
 /**
···
  *
  * Note: Must call it in process context since it might sleep.
  */
-static void subscr_terminate(struct tipc_subscriber *subscriber)
+static void subscr_terminate(struct tipc_subscription *sub)
 {
-	tipc_conn_terminate(&topsrv, subscriber->conid);
+	struct tipc_subscriber *subscriber = sub->subscriber;
+	struct tipc_net *tn = net_generic(sub->net, tipc_net_id);
+
+	tipc_conn_terminate(tn->topsrv, subscriber->conid);
 }
 
 static void subscr_release(struct tipc_subscriber *subscriber)
···
  */
 static int subscr_subscribe(struct net *net, struct tipc_subscr *s,
 			    struct tipc_subscriber *subscriber,
-			    struct tipc_subscription **sub_p) {
+			    struct tipc_subscription **sub_p)
+{
+	struct tipc_net *tn = net_generic(net, tipc_net_id);
 	struct tipc_subscription *sub;
 	int swap;
 
···
 	}
 
 	/* Refuse subscription if global limit exceeded */
-	if (atomic_read(&subscription_count) >= TIPC_MAX_SUBSCRIPTIONS) {
+	if (atomic_read(&tn->subscription_count) >= TIPC_MAX_SUBSCRIPTIONS) {
 		pr_warn("Subscription rejected, limit reached (%u)\n",
 			TIPC_MAX_SUBSCRIPTIONS);
 		return -EINVAL;
···
 	sub->subscriber = subscriber;
 	sub->swap = swap;
 	memcpy(&sub->evt.s, s, sizeof(struct tipc_subscr));
-	atomic_inc(&subscription_count);
+	atomic_inc(&tn->subscription_count);
 	if (sub->timeout != TIPC_WAIT_FOREVER) {
 		setup_timer(&sub->timer, subscr_timeout, (unsigned long)sub);
 		mod_timer(&sub->timer, jiffies + sub->timeout);
···
 	if (subscr_subscribe(net, (struct tipc_subscr *)buf, subscriber,
 			     &sub) < 0) {
 		spin_unlock_bh(&subscriber->lock);
-		subscr_terminate(subscriber);
+		subscr_terminate(sub);
 		return;
 	}
 	if (sub)
 		tipc_nametbl_subscribe(sub);
 	spin_unlock_bh(&subscriber->lock);
 }
-
 
 /* Handle one request to establish a new subscriber */
 static void *subscr_named_msg_event(int conid)
···
 	return (void *)subscriber;
 }
 
-int tipc_subscr_start(void)
+int tipc_subscr_start(struct net *net)
 {
-	return tipc_server_start(&topsrv);
+	struct tipc_net *tn = net_generic(net, tipc_net_id);
+	const char name[] = "topology_server";
+	struct tipc_server *topsrv;
+	struct sockaddr_tipc *saddr;
+
+	saddr = kzalloc(sizeof(*saddr), GFP_ATOMIC);
+	if (!saddr)
+		return -ENOMEM;
+	saddr->family			= AF_TIPC;
+	saddr->addrtype			= TIPC_ADDR_NAMESEQ;
+	saddr->addr.nameseq.type	= TIPC_TOP_SRV;
+	saddr->addr.nameseq.lower	= TIPC_TOP_SRV;
+	saddr->addr.nameseq.upper	= TIPC_TOP_SRV;
+	saddr->scope			= TIPC_NODE_SCOPE;
+
+	topsrv = kzalloc(sizeof(*topsrv), GFP_ATOMIC);
+	if (!topsrv) {
+		kfree(saddr);
+		return -ENOMEM;
+	}
+	topsrv->net			= net;
+	topsrv->saddr			= saddr;
+	topsrv->imp			= TIPC_CRITICAL_IMPORTANCE;
+	topsrv->type			= SOCK_SEQPACKET;
+	topsrv->max_rcvbuf_size		= sizeof(struct tipc_subscr);
+	topsrv->tipc_conn_recvmsg	= subscr_conn_msg_event;
+	topsrv->tipc_conn_new		= subscr_named_msg_event;
+	topsrv->tipc_conn_shutdown	= subscr_conn_shutdown_event;
+
+	strncpy(topsrv->name, name, strlen(name) + 1);
+	tn->topsrv = topsrv;
+	atomic_set(&tn->subscription_count, 0);
+
+	return tipc_server_start(topsrv);
 }
 
-void tipc_subscr_stop(void)
+void tipc_subscr_stop(struct net *net)
 {
-	tipc_server_stop(&topsrv);
+	struct tipc_net *tn = net_generic(net, tipc_net_id);
+	struct tipc_server *topsrv = tn->topsrv;
+
+	tipc_server_stop(topsrv);
+	kfree(topsrv->saddr);
+	kfree(topsrv);
 }

net/tipc/subscr.h

···
 
 int tipc_subscr_overlap(struct tipc_subscription *sub, u32 found_lower,
 			u32 found_upper);
-
 void tipc_subscr_report_overlap(struct tipc_subscription *sub, u32 found_lower,
 				u32 found_upper, u32 event, u32 port_ref,
 				u32 node, int must);
-
-int tipc_subscr_start(void);
-
-void tipc_subscr_stop(void);
+int tipc_subscr_start(struct net *net);
+void tipc_subscr_stop(struct net *net);
 
 #endif